library(keras)
library(tensorflow)
tf$compat$v1$disable_eager_execution()


deprocess_image <- function(x) {
  dms <- dim(x)
  x <- x - mean(x) 
  x <- x / (sd(x) + 1e-5)
  x <- x * 0.1 
  x <- x + 0.5 
  x <- pmax(0, pmin(x, 1))
  array(x, dim = dms)
}

generate_pattern <- function(layer_name, filter_index, size = 150) {
  
  layer_output <- model$get_layer(layer_name)$output
  loss <- k_mean(layer_output[,,,filter_index]) 
  grads <- k_gradients(loss, model$input)[[1]]
  
  grads <- grads / (k_sqrt(k_mean(k_square(grads))) + 1e-5)
  
  iterate <- k_function(list(model$input), list(loss, grads))
  
  input_img_data <- 
    array(runif(size * size * 3), dim = c(1, size, size, 3)) * 20 + 128
  
  step <- 1
  for (i in 1:40) {
    c(loss_value, grads_value) %<-% iterate(list(input_img_data))
    input_img_data <- input_img_data + (grads_value * step) 
  }
  
  img <- input_img_data[1,,,]
  deprocess_image(img) 
}


library(grid)

library(keras)
model <- application_vgg16(
  weights = "imagenet", 
  include_top = FALSE
)
layer_name <- "block3_conv1"
filter_index <- 1
grid.raster(generate_pattern("block3_conv1", 1))